Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

A general perspective on CBO methods with stochastic rate of information

T0 review · 2 major / 4 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper proves that an arbitrarily small positive initial level of information suffices for Consensus-Based Optimization particles to concentrate on the global minimizer in finite time once the concentration parameter is large.

desk verdict A real generalization of CBO convergence with an honest proof, but the abstract overstates the result by hiding a load-bearing spatial-support assumption that is necessary, not cosmetic. read the letter →

arxiv 2507.20029 v1 pith:NLNEENKA submitted 2025-07-26 math.OC math.AP

Stefano Almi , Alessandro Baldi , Marco Morandotti , Francesco Solombrino This is my paper · ORCID
classification math.OCmath.AP MSC 34F0560K3593A1660H1090C5635Q84
keywords Consensus-BasedOptimizationstochasticinformationrateMcKean-VlasovSDEmean-fieldlimitFokker-Planckequationconcentrationtowardsconsensusmulti-agentsystems
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Consensus-Based Optimization (CBO) replaces a swarm of agents exploring an energy landscape with particles that drift toward a consensus point while exploring randomly. This paper treats each agent's "rate of information" $\Lambda_t$ as a time-evolving stochastic label: an informed agent ($\lambda$ near 1) follows the consensus estimate, while an uninformed one follows the crowd average. The main result is that a positive, however small, initial level of information, combined with an initial distribution that puts positive mass in every ball around the minimizer and sufficiently small noise, makes the particles concentrate around the global minimizer in finite time once the concentration parameter $n$ is large. The paper also proves well-posedness, a finite-particle approximation, and mean-field convergence to a kinetic PDE, and shows that the original CBO models fit the same abstract assumptions.

What carries the argument

The machinery is a comparison argument between the original system (4.1) and the auxiliary system (4.2) with $f=0$. In the auxiliary system the velocity is $-X_t + (1-\Lambda_t)e(\mu_t)$, and the key estimate is that under (T3) the average information $E(\Lambda_t)$ stays bounded away from zero, which prevents the information from vanishing too quickly and supplies the divergence condition $\int_0^\infty E(\Lambda_s)\,ds=\infty$. A Grönwall-Itô calculation then forces $E(\|X_t\|^2)\to 0$. To transfer this to the full system, Proposition 4.4 shows that along the evolution the mass in every ball $B_r$ around the minimizer decays at worst exponentially, so with $\mu_0(B_r)>0$ the assumption (f3) makes $f_n(\rho^n_s)\to 0$ uniformly in $s$. A second-moment comparison bound then shows that $X^n_t$ and $X_t$ are close for large $n$.

What would settle it

Run the Gibbs-weighted particle system with a unique minimizer at 0, $\varsigma^2 d<2$, large $n$, and a generator $T$ satisfying (T1)-(T2) but with $T_\Psi(x,0)=0$ on a set of positive measure around 0; if $E(\|X^n_t\|^2)$ remains bounded away from 0 as $t$ grows, assumption (T3) is essential. Conversely, a simulation with full-support initial data, such as a Gaussian centered at 0, and a $T$ satisfying all four hypotheses should reproduce the predicted finite-time concentration, and failure there would indicate an error in the comparison step.

Watch

Extended reading notes

Core claim

The central claim is Theorem 4.5: for the McKean-Vlasov system $dX_t = v^n_{\rho_t}(X_t,\Lambda_t)\,dt + \varsigma\,\|v^n_{\rho_t}(X_t,\Lambda_t)\|\,dB_t$, $d\Lambda_t = T_{\Sigma_t}(X_t,\Lambda_t)\,dt$, under abstract hypotheses (f1)-(f3), (T1)-(T3), the condition $\varsigma^2 d < 2$, $E(\Lambda_0)>0$, and $\mu_0(B_r)>0$ for every $r>0$, for every $\varepsilon>0$ there exist $T_\varepsilon>0$ and $n_\varepsilon\in\mathbb{N}$ such that $E(\|X^n_{T_\varepsilon}\|^2)\le\varepsilon$ for every $n\ge n_\varepsilon$. The proof compares the full system with an auxiliary system whose drift has $f=0$, shows that in the auxiliary system particles collapse to the minimizer because the average information cannot vanish too fast, and then transfers this convergence back through a uniform estimate on $f_n(\rho^n_s)$. Section 5 verifies that the classical Gibbs-weighted CBO drift satisfies the abstract hypotheses, so the result covers the first CBO models proposed in the literature.

Load-bearing premise

The load-bearing premise is that the initial spatial distribution puts positive probability in every ball around the minimizer; if the swarm starts away from the minimizer, the lower mass bound that feeds assumption (f3) fails and the comparison with the auxiliary system breaks.

Editorial extensions

If this is right

  • Concentration in finite time holds for any drift $f_n$ satisfying the abstract axioms, not only for the Gibbs-weighted average of the original CBO model.
  • The classical CBO schemes of [12] and [24] are recovered as special cases, so their global convergence is re-derived from a common toolbox.
  • The finite-particle system inherits the concentration: for large $n$ and $N$, the empirical measure at time $T_\varepsilon$ has second moment below $\varepsilon$ (Corollary 4.6).
  • A positive initial level of information $E(\Lambda_0)>0$ is sufficient; no threshold strength of the information is required, only that it is not exactly zero.
  • The mean-field PDE has a unique solution in $C([0,T]; \mathcal{P}_2(\mathbb{R}^d\times[0,1]))$, giving a kinetic description of the informed swarm.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The condition $\mu_0(B_r)>0$ suggests a design rule: initialization must include exploration across the whole domain, since a swarm seeded only away from the optimum falls outside the theorem's hypotheses and adding a small uniform exploration component to the initial law would restore them.
  • The finite-time nature of the bound raises a natural next question, which the paper does not address: how $T_\varepsilon$ and $n_\varepsilon$ scale with $\varepsilon$ and with the initial information level; testing those scalings numerically is a direct extension.
  • Because $\lambda$ lives in $[0,1]$ and evolves by a jump-type generator, the framework connects to label-switching population dynamics, where $\lambda$ could be read as the probability that an agent belongs to an informed subpopulation.
  • Assumption (T3) is only strict positivity at $\lambda=0$; the proof suggests that any mechanism keeping agents from being fully uninformed for too long would play the same role, so relaxations with $T_\Psi(x,0)$ vanishing on small sets might still yield concentration under modified lower-bound arguments.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper develops a general mean-field framework for Consensus-Based Optimization (CBO) in which each agent carries a stochastic information rate Lambda_t in [0,1]. The dynamics are McKean-Vlasov SDEs with drift assembled from a general consensus-type map f_n(mu), the mean e(mu), and the rate Lambda_t, plus a diffusion whose coefficient vanishes at the consensus point. The authors prove well-posedness by truncation, derive the mean-field PDE and a finite-particle approximation with convergence rates, and then establish the main concentration result (Theorem 4.5): under abstract conditions (f1)-(f3), (T1)-(T3), the small-noise condition sigma^2 d < 2, E(Lambda_0) > 0, and mu_0(B_r) > 0 for every r > 0, for every eps > 0 there exist T_eps and n_eps such that E(||X^n_{T_eps}||^2) <= eps for all n >= n_eps. The proof compares the original system with an auxiliary system with f = 0, whose second moment decays by Theorem 4.1 once E(Lambda_t) does not vanish too fast, and Theorem 4.2 shows the latter follows from (T3). The comparison uses (f3) to make f_n(rho^n_s) -> 0 uniformly, with the lower mass estimate of Proposition 4.4. Section 5 verifies the abstract assumptions for the classical CBO weights.

Significance. If the results hold, the paper provides a useful abstract toolbox that unifies several existing CBO convergence results and extends them to a stochastic, time-evolving information rate. The finite-time concentration statement of Theorem 4.5 is stronger than the usual asymptotic statements and is accompanied by explicit hypotheses that are verified for the classical CBO example in Section 5. The paper also proves finite-particle and mean-field approximation statements, which are important for the practical relevance of the model. A notable strength is that the assumptions (f1)-(f3) and (T1)-(T3) are stated in a way that makes the mechanism of the proof transparent rather than being hidden in a specific Gibbs-weight computation.

major comments (2)
  1. [Abstract and Section 4.2 (Theorem 4.5)] The advertised claim that 'a positive, however small, initial level of knowledge is enough for convergence to consensus' is misleading. Assumption (4.30) requires not only E(Lambda_0) > 0 but also mu_0(B_r) > 0 for every r > 0, i.e. the initial spatial distribution must charge every neighborhood of the minimizer. This condition is load-bearing: it enters through Proposition 4.4, which supplies the lower bound rho^n_t(B_r) >= E(phi_r(X_0)) e^{-q t}, and only then can (f3) be applied to force f_n(rho^n_s) -> 0 uniformly. For the classical CBO model of Section 5 with g(x) = x, the choice mu_0 = delta_{x_0} with x_0 != 0 and E(Lambda_0) > 0 satisfies every hypothesis except (4.30); in that case v^n = 0 and the unique solution is X^n_t = x_0, so the conclusion (4.31) fails. The theorem itself is internally consistent, but the abstract and the introduction should be rewritten to state explicitly that the initial spatial distribution must already have positive mass in every ball around the minimizer, and that this condition is as essential as E(Lambda_0) > 0.
  2. [Section 2, Proposition 2.2] The well-posedness of the generic McKean-Vlasov system (2.1) is imported from the authors' unpublished preprint [5, Theorem 4.2]. This is a load-bearing point, because Theorem 3.3, the mean-field limit, and ultimately Theorem 4.5 all rely on Proposition 2.2. Since [5] is not a published reference, the present paper is not self-contained on this central point. The authors should either include a proof of Proposition 2.2 (or a precise statement with all hypotheses and a full argument) or update the reference to a published version if one becomes available.
minor comments (4)
  1. [Section 4.1, proof of Theorem 4.2] The notation 't := argmin{t in [0,+infinity) : E(Lambda_t) <= delta}' is not mathematically appropriate because an argmin is not defined in this way for a continuous function on an unbounded interval. It should be replaced by 't := inf{t >= 0 : E(Lambda_t) <= delta}' and the continuity argument adjusted accordingly.
  2. [Corollary 4.6] The proof uses convergence of the full sequence Sigma^{n,N}_{T_eps} as N -> infinity, but Lemma 3.11 only establishes convergence along a subsequence. The full convergence follows from the uniqueness of the limit in Proposition 3.13, but this should be stated explicitly.
  3. [Lemma 3.11 and Corollary 4.6] There is a typo in the displayed line 'Sigma^{n,N}_{T_eps/2}' in the proof of Corollary 4.6: the subscript should be T_eps, not T_eps/2.
  4. [Throughout] Several typos should be corrected: 'aknowledges' -> 'acknowledges', 'the the FWF' -> 'the FWF', 'of of our analysis' -> 'of our analysis', 'Unversità' -> 'Università', and 'desiderable' -> 'desirable'.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: Theorem 4.5 is proved by comparison with an f=0 auxiliary system; only a companion-preprint self-citation and an abstract overstatement of hypotheses are noted.

full rationale

The concentration theorem (Theorem 4.5) is not circular. The proof first shows the auxiliary f=0 system satisfies E||X_t||^2 -> 0 (Theorem 4.2, via Grönwall estimates and (T3)), then bounds E||X^n_{Tε} - X_{Tε}||^2 by C∫||f_n(ρ^n_s)||^2 ds and uses (f3) with ℓ(r) = E(ϕ_r(X0)) e^{-qTε} to make this vanish. Assumption μ0(B_r)>0 enters only through Proposition 4.4's lower bound ρ^n_t(B_r) ≥ E(ϕ_r(X0)) e^{-qt}; it is not the conclusion, since a measure can charge every ball around 0 and still have large second moment. Section 5 verifies (f1)-(f3) for the classical CBO weights without invoking the conclusion. The only self-citation is Proposition 2.2, whose proof is deferred to [5, Theorem 4.2] (Baldi-Morandotti, two of the present authors). This is a parameter-free well-posedness statement on a more general system whose assumptions do not include the concentration result, so under the stated rules it is real evidence rather than a circular reduction. The abstract does overstate the theorem: 'a positive, however small, initial level of knowledge is enough' omits the load-bearing condition μ0(B_r)>0, and for μ0=δ_{x0} with x0≠0 the standard CBO dynamics is stationary, so the conclusion fails. This is a scope mismatch, not circularity. A footnote in Proposition 3.10 also signals an omitted stochastic-stopping-time technical detail, a completeness issue rather than a circular step.

Assumptions & free parameters 0 free parameters · 8 assumptions · 1 invented entities

The central claim rests on the stated axioms for f_n and T, the noise bound ς^2 d < 2, the initial-data assumptions, and the well-posedness result imported from the authors' preprint [5]. No parameters are fitted to data; the only new entity is the stochastic information-rate variable Λ_t, which is a modeling construct without independent empirical evidence.

assumptions (8)
  • ad hoc to paper Existence and pathwise uniqueness for the generic McKean-Vlasov system with Lipschitz coefficients (Proposition 2.2, imported from [5, Theorem 4.2])
    Proposition 2.2 rests on the authors' preprint [5]; the proof is not included in this paper, making it a load-bearing external dependency.
  • domain assumption (f1)-(f2): f_n is locally Lipschitz in W1 and has linear growth with uniform constant M
    Assumptions on f_n in Section 3, used to prove well-posedness, moment bounds, and the truncation argument.
  • domain assumption (f3): f_n converges to 0 uniformly on sets of measures with bounded second moment and a positive lower bound on mass near 0
    Core assumption for the comparison with the auxiliary system in Theorem 4.5; verified for classical CBO in Proposition 5.3.
  • domain assumption (T1)-(T3): T is Lipschitz, λ + θ T_Ψ(x,λ) ∈ [0,1], and T_Ψ(x,0) > 0
    Used to prove the positivity of the averaged information rate and the convergence of the auxiliary system in Theorem 4.2.
  • domain assumption ς^2 d < 2
    Required in Theorem 4.1 and Theorem 4.2 for the exponential decay of the second moment in the auxiliary system.
  • domain assumption Initial data: X0 ∈ L4, E(Λ0) > 0, μ0(B_r) > 0 for every r > 0
    Assumption (4.30) in Theorem 4.5; encodes the 'positive initial knowledge' and the spatial support condition.
  • domain assumption For the standard CBO application: E has unique minimizer at 0, coercivity and growth conditions (5.4)-(5.7), g Lipschitz or bounded
    Proposition 5.1 and Proposition 5.3 verify (f1)-(f3) for the Gibbs-weighting case under these conditions.
  • standard math Standard Itô calculus, Grönwall lemma, Aldous criterion, Skorokhod representation theorem, Vitali convergence theorem
    Background tools used throughout the proofs without further justification.
invented entities (1)
  • Λ_t: stochastic information rate of each agent
    purpose: Models each agent's fluctuating knowledge of the energy landscape; enters the drift as a convex combination between the consensus point and the mean position, and gates the noise intensity.
    Introduced as a new state variable in system (1.5); no empirical calibration or falsifiable prediction outside the model is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A general perspective on CBO methods with stochastic rate of information." pith.science (2026). https://pith.science/paper/NLNEENKA

@misc{pith2026250720029,
  author       = {Pith},
  title        = {Pith review of: A general perspective on CBO methods with stochastic rate of information},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NLNEENKA}},
  note         = {Machine review of arXiv:2507.20029}
}
read the original abstract

This paper studies a class of Consensus-Based Optimization (CBO) models featuring an additional stochastic rate of information, modeling the agents' knowledge of the environment and energy landscape. The well-posedness of the stochastic system is proved, together with its finite-particle approximation and the mean-field convergence to a kinetic PDE. Particles are shown to concentrate around the consensus point under mild assumptions on the initial spatial distribution and initial level of knowledge. In particular, the analysis unveils that a positive, however small, initial level of knowledge is enough for convergence to consensus to happen. The framework presented is general enough to include the first instances of CBO proposed in the literature.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. An alternative approach to well-posedness of McKean-Vlasov equations arising in Consensus-Based Optimization

    math.OC 2025-12 accept novelty 5.0 of 10

    A truncation argument on the Wasserstein space yields existence and pathwise uniqueness for the mean-field CBO equation, extending uniqueness to solutions whose consensus point is bounded rather than continuous.

Reference graph

Works this paper leans on

43 extracted references · 42 canonical work pages · cited by 1 Pith paper

  1. [12]

    J. A. Carrillo, Y-P. Choi, C. Totzeck, and O. Tse, An analytical framework for consensus-based global optimization method, Math. Models Methods Appl. Sci. 28 (2018), no. 6, pp. 1037–1066

    work page 2018

  2. [24]

    Fornasier, T

    M. Fornasier, T. Klock, and K. Riedl, Consensus-based optimization methods converge globally. SIAM J. Optim. 34 (2024), pp. 2973–3004

    work page 2024

  3. [5]

    Baldi and M

    A. Baldi and M. Morandotti, Well-posedness and propagation of chaos for multi-agent models with strategies and diffusive effectsPreprint arxiv.org/abs/2507.14058

    arXiv

  4. [1]

    Aarts and J

    E. Aarts and J. Korst, Simulated annealing and Boltzmann machines. A stochastic approach to combinatorial optimization and neural computing. Wiley-Interscience Series in Discrete Mathematics and Optimization. John Wiley & Sons, Ltd., Chichester, 1989

    work page 1989

  5. [2]

    Ambrosio, M

    L. Ambrosio, M. Fornasier, M. Morandotti, and G. Savaré, Spatially inhomogeneous evolutionary games, Comm. Pure Appl. Math., 74 (2021), pp. 1353–1402

    work page 2021

  6. [3]

    T. Back, D. B. Fogel, and Z. Michalewicz, Handbook of evolutionary computation. IOP Publishing Ltd., 1997

    work page 1997

  7. [4]

    Baldi, Stochastic calculus

    P. Baldi, Stochastic calculus. An introduction through theory and exercises, Springer, Cham, 2017

    work page 2017

  8. [6]

    Benamou, G

    J.-D. Benamou, G. Carlier, M. Cuturi, L. Nenna, and G. Peyré, Iterative Bregman projections for regularized transportation problems, SIAM J. Sci. Comput., 37 (2015), pp. A1111–A1138

    work page 2015

Show all 43 references
  1. [7]

    Billingsley, Convergence of probability measures, Wiley Series in Probability and Statistics, 2nd Edition, 1999

    P. Billingsley, Convergence of probability measures, Wiley Series in Probability and Statistics, 2nd Edition, 1999

    1999

  2. [8]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi, An adaptive consensus based method for multiobjective optimization with uniform Pareto front approximation, Appl. Math. Optim., 88 (2023), paper n. 58

    2023

  3. [9]

    Borghi, M

    G. Borghi, M. Herty, and L. Pareschi, Constrained consensus-based optimization. SIAM J. Optim. 33 (2023), pp. 211–236

    2023

  4. [10]

    H. Brézis, Opérateurs maximaux monotones et semi-groupes de contractions dans les espaces de Hilbert, North- Holland Publishing Co., Amsterdam-London; American Elsevier Publishing Co., Inc., New York, 1973

    1973

  5. [11]

    J. A. Cañizo, J. A. Carrillo, and J. Rosado, A well-posedness theory in measures for some kinetic models of collective motion, Math. Models Methods Appl. Sci. 21 (2011), pp. 515–539

    2011

  6. [13]

    J. A. Carrillo, S. Jin, L. Li, and Y. Zhu, A consensus-based global optimization method for high dimensional machine learning problems. ESAIM Control Optim. Calc. Var., 27 (2021), paper n. S5

    2021

  7. [14]

    J. A. Carrillo, C. Totzeck, and U. Vaes, Consensus-based optimization and ensemble kalman inversion for global optimization problems with constraints. Modeling and Simulation for Collective Dynamics, pp. 195–230. World Scientific, 2023

    2023

  8. [15]

    Cheng, N

    X. Cheng, N. Chatterji, P. Bartlett, and M. Jordan, Underdamped Langevin MCMC: A non-asymptotic analysis, in Proc. Conf. on Learning Theory, 2018, pp. 300–323

    2018

  9. [16]

    Dembo and O

    A. Dembo and O. Zeitouni, Large Deviations Techniques and Applications, volume 38. Springer Science & Business Media, 2009

    2009

  10. [17]

    Dupuis and R

    P. Dupuis and R. S. Ellis, A Weak Convergence Approach to the Theory of Large Deviations, Wiley, New York, 1997

    1997

  11. [18]

    Düring, P

    B. Düring, P. Markowich, J.-F. Pietschmann, and M.-T. Wolfram, Boltzmann and Fokker-Planck equations modelling opinion formation in the presence of strong leaders, Proc. R. Soc. Lond., Ser. A, Math. Phys. Eng. Sci. 465 (2009), pp. 3687–3708

    2009

  12. [19]

    D’Onofrio and A

    G. D’Onofrio and A. M. Hernandez, A large multi-agent system with noise both in position and control, preprint arxiv.org/abs/2503.10543

    arXiv

  13. [20]

    D. B. Fogel, Evolutionary computation. Toward a new philosophy of machine intelligence, IEEE Press, Piscataway, NJ, second edition, 2000

    2000

  14. [21]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. Sünnen, Consensus-based optimization on hypersurfaces: Well- posedness and mean-field limit, Math. Models Methods Appl. Sci. 30 (2020), pp. 2725–2751

    2020

  15. [22]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. Sünnen, Consensus-based optimization on the sphere: convergence to global minimizers and machine learning, J. Mach. Learn. Res. 22 (2021), paper n. 237, pp. 1–55

    2021

  16. [23]

    Fornasier, H

    M. Fornasier, H. Huang, L. Pareschi, and P. Sünnen, Anisotropic diffusion in consensusbased optimization on the sphere, SIAM J. Optim. 32 (2022), pp. 1984–2012

    2022

  17. [25]

    Fornasier and F

    M. Fornasier and F. Solombrino, Mean-field optimal control, ESAIM Control Optim. Calc. Var. 20 (2014), pp. 1123–1152

    2014

  18. [26]

    Fornasier and L

    M. Fornasier and L. Sun, A PDE framework of consensus-based optimization for objectives with multiple global minimizers, Comm. Partial Differential Equations 50 (2025), no. 4, pp. 493–541

    2025

  19. [27]

    Grassi and L

    S. Grassi and L. Pareschi, From particle swarm optimization to consensus based optimization: stochastic modeling and mean-field limit, Math. Models Methods Appl. Sci., 31 (2021), pp. 1625–1657

    2021

  20. [28]

    J. H. Holland, Adaptation in natural and artificial systems. An introductory analysis with applications to biology, control, and artificial intelligence, University of Michigan Press, Ann Arbor, Mich., 1975

    1975

  21. [29]

    Huang and J

    H. Huang and J. Qiu, On the mean-field limit for the consensus-based optimization, Math. Methods Appl. Sci. 45 (2022), pp. 7814–7831

    2022

  22. [30]

    Kennedy and R

    J. Kennedy and R. Eberhart, Particle swarm optimization, in Proc. IEEE Int. Conf. Neural Networks, 1995, pp. 1942–1948

    1995

  23. [31]

    Kirkpatrick, C

    S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, Optimization by simulated annealing, Science, 220 (1983), pp. 671–680

    1983

  24. [32]

    Klamroth, M

    K. Klamroth, M. Stiglmayr, and C. Totzeck, Consensus-based optimization for multi-objective problems: a multi-swarm approach, J. Global Optim. 89 (2024), no. 3, pp. 745–776

    2024

  25. [33]

    Le Bris and P.-L

    C. Le Bris and P.-L. Lions, Existence and Uniqueness of Solutions to Fokker-Planck Type Equations with Irregular Coefficients, Communications in Partial Differential Equations, 33 (2008), pp. 1272–1317

    2008

  26. [34]

    Loy and A

    N. Loy and A. Tosin, Boltzmann-type equations for multi.agent systems with label switching, Kinet. Relat. Models 14 (2021), n. 5, pp. 867–894

    2021

  27. [35]

    Morandotti and F

    M. Morandotti and F. Solombrino, Mean-field Analysis of Multipopulation Dynamics with Label Switching, SIAM J. Math. Anal., 52 (2020), pp. 1427–1462. A GENERAL PERSPECTIVE ON CBO METHODS WITH STOCHASTIC RATE OF INFORMATION 25

    2020

  28. [36]

    Øksendal, Stochastic differential equations, Universitext, Springer-Verlag, Berlin, sixth ed., 2003

    B. Øksendal, Stochastic differential equations, Universitext, Springer-Verlag, Berlin, sixth ed., 2003. An introduction with applications

    2003

  29. [37]

    G. A. Pavliotis, Stochastic Processes and Applications, Springer, New York, 2014

    2014

  30. [38]

    Piccoli and F

    B. Piccoli and F. Rossi, Measure-theoretic models for crowd dynamics, Crowd dynamics. Vol. 1, Model. Simul. Sci. Eng. Technol., Birkhäuser/Springer 2018,

    2018

  31. [39]

    Pinnau, C

    R. Pinnau, C. Totzeck, O. Tse, and S. Martin, A consensus-based model for global optimization and its mean-field limit, Math. Mod. Meth. Appl. Sci., 27 (2017), pp. 183–204

    2017

  32. [40]

    Raginsky, A

    M. Raginsky, A. Rakhlin, and M. Telgarsky, Non-convex learning via stochastic gradient Langevin dynamics: a nonasymptotic analysis, in of the 2017 Conference on Learning Theory, PMLR 65, 2017, pp. 1674–1703

    2017

  33. [41]

    Storn and K

    R. Storn and K. Price, Differential evolution – A simple and efficient heuristic for global optimization over continuous spaces, J. Global Optim., 11 (1997), pp. 341–359

    1997

  34. [42]

    Toscani, Kinetic models of opinion formation, Commun

    G. Toscani, Kinetic models of opinion formation, Commun. Math. Sci. 4 (2006), pp. 481–496

    2006

  35. [43]

    R. Caccioppoli

    C. Villani, Optimal Transport: Old and New, Springer Berlin, Heidelberg 2008. (Stefano Almi)Dipartimento di Matematica e Applicazioni “R. Caccioppoli”, Unversità di Napoli “Federico II”, Via Cintia, 80126 Napoli, Italy. ORCID: 0000-0001-7308-221X. Email address: stefano.almi@u...

    2008

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.